"trackeff" module line improvements for August MD

This MR:

• removes the IP cut from the calibration lines (lines with the full raw event) that we have for the Z and upsilon. After discussion, we decided that it would be useful to remove these to make debugging possible more easily. Lines are used for the calibration of the tracker's hit efficiency and the detector alignment. To avoid the QEE maintainers the extra work of maintaining these lines, I moved them to the trackeff stream as well, as these are under the scrutiny of the WP4 conveners.
• renamed the SelectionFromWeightedRelations to SelectionFromRelations, as it didn't work on weighted relations
• enables selective persistence for the KS -> pi pi velo2long lines, and cut their rates by about 15% - the idea is that we scrutinize this selective persistence in the next runs, and in the next round of line changes, we disable the persistence of all tracks. Expect that this reduces the bandwidth by 50%
• adds lines for Ds -> phi pi with phi -> KK for velo2long efficiencies, directly using the selective persistence. These are less critical but would be desirable for the studies of the material map. I added a prescale to get the rate to 200Hz.

Rates (based on 400k HLT1-filtered data with mu=3.5)

Decay Mode	Before	After
Z0	< 5 Hz	< 5 Hz
Upsilon	55 Hz	42.5 Hz
KS0 (all)	1000 Hz	850 Hz
Ds -> Phi pi	N/A	200 Hz

I have checked on a simulation sample that the Z line is selecting events. Altogether, I expect that the difference in the overall stream bandwidth is small, with a few lines going down but a few new things added.

Goes with: Rec!4019 (merged)

FYI @poluekt @freiss